Tanzania Journal of Science 44(3): 31-45, 2018 ISSN 0856-1761, e-ISSN 2507-7961 © College of Natural and Applied Sciences, University of Dar es Salaam, 2018

APPLICATION OF REGRESSION MODEL TO IDENTIFY A PARAMETER THAT BEST DEFINES SPECIES DIVERSITY IN THE COASTAL FORESTS OF TANZANIA

Cosmas Mligo Department of Botany, P.O. Box 35060, Botany Department, University of Dar es Salaam, Tanzania. [email protected], [email protected]

ABSTRACT Coastal forests of Tanzania are diverse in plant species that make them included as part of the 34 world biodiversity hotspots. This study aimed at determining plant species diversity, richness, and evenness as well as to identify the parameter that best defines plant species diversity in the three coastal forests; namely Zaraninge, Kazimzumbwi and Pande. Transect method was used for vegetation data collection and subjected to Analysis of variance and regression techniques. The plant species composition among forests ranged between 75 and 146 with a significantly lower diversity index in Zaraninge (2.057 ± 0.112) than those in Pande (2.415 ± 0.022) and Kazimzumbwi (2.578 ± 0.092) forests. The plant species were more evenly distributed in Zaraninge (0.488 ± 0.004) than in Pande (0.452 ± 0.016) and Kazimzumbwi (0.457 ± 0.025) with no significant difference among them. The species richness per plot was significantly lower in Zaraninge forest (14 ± 1) than Kazimzumbwi forest (20 ± 2). Data on evenness and richness were correlated with plant diversity indices at different levels. A perfect positive correlated occurs with evenness (r =1) but lower with richness (Zaraninge, r = 0.88; Pande, r = 0.91 and r = 0.79 for Kazimzumbwi). This implies that richness and evenness parameters portray different ecological interpretations of the biodiversity value within an ecosystem and cannot be used interchangeably. Regression models showed that species evenness significantly determined plant species diversity, whereas richness was not significant. This study concludes that evenness stands a better chance of being the best predictor of change in plant species diversity and therefore an adequate measure of the coastal forests’ conservation value than richness.

Key words: Coastal forest, conservation, diversity, evenness, richness, regression model

INTRODUCTION as part of the 34 world biodiversity hotspots Coastal forests of Tanzania are part of the (Myers 2002). Plant species surviving within East African coastal forest belt covering the fragmented habitats or localized within a southern coast of Kenya and throughout the single forest are regarded as paleo-endemic coastal strip of Tanzania including the due to long term isolation from other forests in the Islands of Zanzibar and Pemba population within an ecosystem. Because of and the northern parts of Mozambique high species diversity characterized by high (Burgess et al. 2000, WWF 2002). These levels of endemism, an appropriate forests are characterized by a mosaic but interpretation of plant species diversity localized habitat types (Burgess and Clarke needs to be established to form the basis of 2000), that have prevented a wide range of conserving the East African coastal forests’ plant species distribution. The presence of ecosystem. species with habitat restrictions makes coastal forests of East Africa being included 31 www.tjs.udsm.ac.tz www.ajol.info/index.php/tjs/

Mligo - Application of regression model to identify a parameter that best defines species

Species diversity and abundance are among parameters. Clarke and Dickinson (1995) the most important themes in ecology and reported that coastal forests are diverse in biodiversity conservation (Watanabe and plant species; this general conclusion was Suzuki 2008). However, there is lack of an based on species richness encountered per appropriate meaning of species diversity and field surveyed. In semi-arid ecosystems in its ecological interpretation at different Tanzania, where livestock grazing pressure habitats (Kent and Coker 1992), making the influence plant species diversity, the species concept hard to apply (Magurran 2004). The evenness becomes the most important than definition of plant species diversity has two species richness (Mligo 2006). Balanced basic components: species richness which is ecosystems (such as Serengeti National the number of plant species and the plant Parks) where plant-animal interactions are in species evenness or equitability which is the a dynamic equilibrium, species diversity is way individuals of each species are defined by both number of species and distributed within a given plant community relative abundance (Mligo 2015). Which (Kent and Coker 1992, Magurran 2004, ecological parameters “either species Hamilton 2005). Efforts have been made to richness or evenness” best defines species establish indices that can be used to describe diversity in the coastal forests of Tanzania. the biodiversity conservation importance of The efforts to conserve coastal forests with an ecosystem (Bock et al. 2007). The large number of species restricted in species richness, evenness and various specialized habitats can be well placed measurements of abundance and rarity are provided an appropriate definition and key components of species diversity, which interpretation of plant species diversity are describe the conservation value of a established. The purpose of this study was to particular ecosystem (Gaston and Spicer determine the plant species diversity, 1998). A number of ecological studies richness and evenness, and identify the established “the plant species richness” as a suitable parameters that provide the best criterion of how an ecosystem is biologically definition and interpretation of plant species diverse (Clarke and Dickson 1995). This is diversity. The assumption was that richness because species richness is the easiest and evenness each provides the same biodiversity parameter to quantify and interpretation of plant species diversity in frequently used as a measure for coastal forests of Tanzania. conservation considerations in the coastal forests of Tanzania. However, diversity in MATERIAL AND METHODS terms of relative abundance (evenness) or Location of the study area spatial patterns among individuals of the Three coastal forests namely; Pande, species in various localities is less Kazimzumbwi and Zaraninge Forests were considered in many studies. Stirling and selected for the purpose of this study (Figure Wilsey (2001) and Ma (2005) pointed out 1). These forests are found between that species richness alone could be a longitudes 380 30’ to 390 6’E and latitudes 50 misleading indicator of phytodiversity 40’ to 70 0S’. To the east, the forests border conservation perspective if it does not the Indian Ocean; to the north is Tanga include other key attributes of species region, the Lindi region to the south and assemblages. Describing the species Morogoro region to the west. The climate is diversity using a single value compromises one of the most important natural influences much of the details of the plant community on the vegetation structure and growth in the structure (Hamilton 2005). Biodiversity coastal area (Clarke and Burgess 2000). The assessment in coastal forests ecosystem in climate of the coastal area is largely tropical, Tanzania may have benefited from similar though some of the southern areas are

32

Tanz. J. Sci. Vol. 44(3) 2018 almost subtropical. There are two rainy amount being 500 millimeters/annum seasons where the long rains occur between (Frazier 1993). The highest average annual April and June and short rains occur rainfall ever recorded in these coastal forests between November and December in the is 1100 millimeters/year (Burgess and north, which may be described as a bimodal Clarke 2000) with temperatures increasing rainfall pattern. The highest amount of from 21.70C in July to 38.50C in November rainfall received in these coastal forests is through December (White 1983). 2,000 millimeters/annum and the lowest

Figure 1: Location of Zaraninge, Pande and Kazimzumbwi Forests in Tanzania

33

Mligo - Application of regression model to identify a parameter that best defines species

Vegetation sampling procedures in the study Where pi = ni/N and is the proportion of the forests total number of all species in a quadrat and After a comprehensive preliminary survey to ln = natural logarithm to base e. identify the existing habitats in each of the three selected coastal forests, sampling of Evenness (E) was calculated using the plant species was carried out. Transects were formula by Alatalo (1981): laid out from the boundary towards the H i Evenness (E)  .....(eq2) forest interior to determine the gradient of ln S change in plant communities. Six transects Where H' is the Shannon-Weaver diversity of 0.65 km long each were established in index and S is the total number of species at each forests, making a total of 18 transects a site. These indices were then compared that covered an area of 1.8 ha in each forest. among forests using Analysis of Variance A total of 36 nested plots as recommended (GraphpadInstat 2003). by Stohlgren et al. (1995) were systematically established after every 100 m Algorithmic establishment of linear along a given transect. The sampling plots regression model were positioned on alternating sides of The model established from the ecological transects following the method of Kasenene parameters determined from vegetation raw (1987). Three levels of sampling were data set of 108 samples from the three employed in the field: (a) A plot of size 25 coastal forests. The parameters (diversity, m × 20 m was used measuring trees with richness and evenness indices) determined >30 cm circumference at the height of 1.3 m from raw data were fit in the observed data from the ground level; (b) 5 m × 2 m to obtain a linear regression model. The subplots nested in the major plot for model was optimized step by step until the assessment of shrubs and saplings; and (c) 2 best fit for the linear relationships among m × 0.5 m subplots nested in the 5 m × 2 m parameters was obtained. subplots for herbs and grasses. A composite data was generated by including all The linear model had only one dependent taxonomic specifications. All plants were variable (Y) and two independent variables identified to species level in the field and for (Xkth). The dependent variable (Y) (i.e. plants that were difficult to identify in the species diversity) was assumed to have field voucher specimens were collected by values above zero from multiple matching with preserved herbarium independent variables. The parameters that specimens of the Department of Botany, had an apparent influence on the species University of Dar Es Salaam. diversity index were regarded as

independent variables (x). The regression Determination of species diversity and model was fit to multiple responses of Y evenness indices from vegetation parameters (evenness and Plant species diversity was determined in richness) that could indicate the change in terms of Shannon diversity index (Shannon plant species diversity of coastal forests. For and Wiener 1948) according to the each possible set of values of the following formula below: independent variables there was a

 probability that a variable significantly Diversity Index (H')   piln pi...(1) influences change in species diversity within i th a plant community. However, the k coefficient was also estimated to show changes in explanatory variables within 34

Tanz. J. Sci. Vol. 44(3) 2018 vegetation data. It was considered internal variation within vegetation data among RESULTS sampling points that made it possible to The plant species composition, diversity, report all the parameters as mean and evenness and richness standard error. The weighted aggregate The difference in plant species composition variation computed delineated the exists among selected coastal forests was magnitude of coefficients between because of different levels of management explanatory variables to identified variables regimes. A total of 75 plant species recorded that correctly defined plant species diversity in Zaraninge forest where Brachylaena in the coastal forests of Tanzania. The huillensis, Croton macrostachyus, Rinorea species diversity was considered to be direct arborea, Haplocoelum mombasense, proportional to the number of plant species Panicum heterostachys and Combretum (richness) H! r and the extent to which harrisii were highly represented in individuals of the species are distributed woodlands, the closed canopy plant ! ! (evenness) i.e. H  e. Therefore H = β2*r community was represented by Cynometra ! and H = β1*e but β1 >0 and β2>0. The webberi, Cynometra brachyrrhachis, species richness (r) was replaced by variable Landolphia kirkii, Scorodophloeus fischeri, X2 and evenness index (e) was replaced by X1. Dalium holztii and Rothmania fischeri. It The decision to predict y = 0 when x2 is less was recorded a total of 120 plant species in than one (x2 <1) or y>0 when x2>1 and x1 >0 Pande Forest where Afzelia quanzense, was based on the linear equation 3. Trema orientalis, Mimusopsis fruitcosa, Dalbergia melanoxylon, Spirostachys kth y = b0 + β1x1 + β2x2 +… βnx … (Equ. 3) africana, Vitex doniana, Colla clavata and Where Y is the logistic transformation of Panicum trichocladum dominated in the species data, which is equivalent to the woodlands. Scorodophloeus fischeri, Salacia predictive Shannon’s diversity index (H!) madagascariensis, Cynometra webberi, x = Vegetation parameters (Evenness -x1 Albizia gumifera, Milletia usambarensis and and Richness-x2) Uvaria kirkii were common in the closed β1 = coefficients of the species evenness canopy plant community. Also a total of 146 (Slope 1) plant species were recorded in β2 =coefficients of the species richness Kazimzumbwi Forest which was represent (Slope 2) by Canthium mombazense, Manilkara b0 = Regression constant sulcata, Grewia platyclada, Strychnos The algorithmic development of the panganiense, Cussonia zimmermannii, regression model was done by using the Mimosa pigra, Albizia petersiana, Milletia Graphpad-Instat statistical software, version impressa, Antiaria toxicaria and Trema 3.06 (GraphpadInstat 2003). The critical orientalis. values were computed as normal F-ratio to determine whether or not pairs of parameters A significant difference in species diversity were significantly related. The coefficients exists among studied forests (P<0.05) (Table of the regression models indicated the 1), and a significantly lower diversity was magnitude of impact of an independent observed in Zaraninge Forest than Pande variable to the dependent variable. The Forest (LSD = 0.520, P<0.01) and coefficient of determination of Kazimzumbwi Forest (LSD = 0.35, P<0.05). multicollinearity was estimated at r2 = 0.75. Similarly, the plant species richness was The graphic patterns of vegetation significantly lower in Zaraninge Forest than parameters were developed based on the in Kazimzumbwi Forest (LSD= 6.86, P = same software (Graphpad Instat 2003). 0.0010) and Pande Forest (LSD = 5.44, P

35

Mligo - Application of regression model to identify a parameter that best defines species

<0.01). However, no significant difference among these forests (P>0.05) (Table 1). was observed in plant species evenness

Table 1: The plant species diversity, evenness and richness within the studied forests (Mean ± SE) and the probability at 5% critical value

Parameters Pande Kazimzumbwi Zaraninge Probability value Shannon-Wiener diversity a2.415 ± b2.578 ± 0.092 c2.057 ± 0.112 P<0.05 index 0.022 Evenness index a 0.488 ± a 0.452 ± 0.016 a 0.457 ± 0.025 P>0.05 0.004 Simpson index a 9.677 ± b 16.75 ± 2.18 c 6.54 ± 0.83 P<0.05 0.305 Richness a 19 ± 1 b 20 ± 2 c 14 ± 1 P<0.05 Different letters (a, b, c) means significant difference and same letters (a, a, a) mean no significant difference

Application of regression model to predict regression models. Within the linear models change in species diversity (Table 2), the signs of the estimated The generated linear regression model was a gradients give the direction of effects of the sufficient predictor of species diversity (y) explanatory variable on the probability of in the coastal forests with varying change in diversity index at any one explanatory variables (xKth) (Table 2). The particular observation. Positive gradients in evenness (X1) contributed more to the evenness (x1) and the richness (x2) suggest change in plant species diversity bearing the that the increase in coefficient of an higher gradient in first part of the regression independent variable gives a high model, than richness that had insignificant probability of its contribution to the species influence. The goodness of fit of the diversity in the forests. The negative regression showed that species evenness estimated gradient on the species richness in significantly influenced change of species Zaraninge Forest showed that species diversity in Kazimzumbwi (F = 3.839 x 109, diversity was determined by species DF = 33, P<0.0001), Zaraninge (F = 2.21 x evenness alone and the dominance by a few 109, DF = 33, P<0.0001) and Pande Forest species is the cause on low richness and (F = 2.619 x 109, DF = 33, P<0.0001). The diversity indices. influence of evenness was consistent and informative on both full and simple

Table 2: The linear models that best fit between diversity (Y) and the determining variables (x1- Evenness and x2-Richness)

Forest N Regression model P Conclusion Zaraninge 36 -7 -5 <0.0001 Significant y = 4.50 x1 – 2.572 × 10 x2 + 2.076 × 10 Pande 36 y = 4.94 + 0.004 × 10-6 + 2.633 × 10-5 <0.0001 Significant Kazimz. 36 -7 -6 <0.0001 Significant y = 5.69 + 4.489 × 10 + 5.398 × 10

36

Tanz. J. Sci. Vol. 44(3) 2018

Based on the linear models, the basic (r = 0.88) and Pande (r = 0.91), decreasing multicolinearity test generated produced to 0.79 for Kazimzumbwi forest (Figure 3). problems with data from Zaraninge and Since the same correlation coefficient values Pande Forests because of low level of obtained when both diversity and evenness disturbance such that richness measures were compared with richness, it implies that contributed redundant information. The the two ecological parameters (richness and direct multicolinearity exists between evenness) conveyed different ecological predictors (either richness or evenness) and interpretation and cannot be used diversity in Kazimzumbwi Forest because of interchangeably to describe the biodiversity being in a heavily disturbed condition (Table value of the coastal forests ecosystem. The 3). The goodness of fit of the model shows trend line on the graphs generated from that, evenness explained 100% of the among parameters for data from Zaraninge variation in species diversity among selected Forest begins as polygonal sharp peaks and coastal forests, whereas richness explained valleys where plots 1-13 showed similar 79.01% in Zaraninge Forest, 83.16 % in pattern (Figure 3). However the pattern Pande Forest and 62.79% in Kazimzumbwi changed between plot 8-10 and between 14- Forest which is less than the minimum 24 as well plots 29-36. The trend line for the percentage coefficient of determination diversity curve had similar pattern with that (75%) for the entire vegetation data set of the species evenness but deviated from (Table 3). The species diversity was perfect the species richness curve (Figure 3). positive correlated with evenness (r =1), but somewhat lower with richness in Zaraninge

37

Mligo - Application of regression model to identify a parameter that best defines species

40 y = 59.476x - 6.4522 40 y = 10.446x - 6.4522 2 2 35 R = 0.6279 35 R = 0.6279 30 30 25 25

20 20 Richness 15 Richness 15 10 10 5 5 0.2 0.3 0.4 0.5 0.6 0.7 1 2 3 4 Evenness (J) Diversity (H!) 0.7 y = 0.2222x - 5E-06 40 y = 40.906x - 5.0877 40 y = 9.0907x - 5.0879 2 2 2 (B) R = 1 35 R = 0.7901 R = 0.7901 0.6 35 30 30 0.5 25 25 0.4 20 20

Evenness (J) Evenness 15 15 Rciness 0.3 Richness 10 10 0.2 5 5 1 2 3 4 0.2 0.3 0.4 0.5 0.6 0.7 1 2 3 Diversity (H!) Evenness (J) Diversity 0.7 y = 0.2024x - 3E-06 y = 58.106x - 9.3387 40 y = 11.759x - 9.3389 (C) 2 2 R = 1 35 R = 0.8428 35 2 0.6 R = 0.8428 30 0.5 25 25 0.4 20

Evenness (J) Evenness 15 15 Richness 0.3 Richness 10 0.2 5 5 1 2 3 4 0.2 0.3 0.4 0.5 0.6 0.7 1 2 3 4 Diversity (H!) Evenness (J) Diversity (H!)

Figure 2: Correlation among ecological parameters (species diversity, richness and Evenness) determined from vegetation data from coastal forests in Tanzania (A) Kazimzumbwi forest (B) Zaraninge Forest and C) Pande Forest.

38

Tanz. J. Sci. Vol. 44(3) 2018

Figure 3: The similarities among trend lines for evenness, richness and diversity indices on vegetation data from Zaraninge Forest Reserve. The evenness trend line is equal to the pattern of species diversity whereas that of richness deviates from the other two. This pattern shows that evenness is an important parameter that provides the best interpretation of species diversity in coastal forests of Tanzania.

39

Mligo - Application of regression model to identify a parameter that best defines species

Table 3: t-test for multicollinearity of the variables (richness and evenness) against variation in species diversity in forests

Site Statistic Evenness Richness Y-intercept Zaraninge t-ratio 3047 0.08 0.06 p-value <0.0001 0.93 0.56 r2 1 0.7901 Remarks Significant not significant not significant Pande t-ratio 28693 0.37 0.62 p-value <0.0001 0.71 0.54 r2 1 0.8316 Remarks Significant not significant not significant Kazimzumbwi t-ratio 53450 0.31 0.17 p-value <0.0001 0.75 0.86 r2 1 0.6279 Remarks Significant not significant not significant

DISCUSSION high values of diversity indices, which can Plant species diversity in the coastal forests mislead the prioritization of biodiversity of Tanzania sanctuary for conservation. Similarly, when It was observed the highest diversity in data from one coastal forest fragment is Kazimzumbwi followed by Pande and considered, species diversity can be lower Zaraninge Forests and the difference was than data from two or more patches or significant among forests studied (P<0.05), multiple forest fragments provided they have (Table 3). Kent and Coker (1992) pointed been collected by using the same sampling out that most Shannon’s diversity indices in techniques. This is because of variation in forest ecosystems ranges between 1.5 and species composition among coastal forests 3.5. Studies in Taita Hill Forests using Point that have high proportion of plant species Centered Quarter method (Wilder et al. with limited distribution pattern (endemism). 1998) reported species diversity in a ranged Variations in environmental variables that between 1.68 and 2.75 from heavily and determine species diversity provide room to least disturbed parts respectively. The compare the same among forests. The diversity indices (with a range between 2.05 habitats in coastal forest ecosystem are and 2.58), obtained in the studied forests fragmented, heterogeneous with soils are indicate that they are moderately rich in predominantly sandy and infertile because of plant species. The diversity indices above potential leaching (Koné et al. 2014). Also were determined from data that covered the climatic conditions in coastal areas are different levels of taxonomic specifications different from other ecosystems due to the because some plant species in forests exists influence of the Indian Ocean of which these as herbs, grass, lianas and shrubs that never forests are in close proximity (Burgess and grow into trees. When a single life form Clarke 2000). So comparing diversity from such as tree or herb is considered separately, the coastal forest ecosystem with other species diversity becomes lower than a ecosystems should be done cautiously. composite data with all life forms inclusive (Mligo 2010). Heyer (1994) pointed out that Among the three forests studied, some tropical forests are species poorer than Kazimzumbwi and Pande Forests are others when only dominant life form is anthropogenically disturbed due to illegal considered. Using such data will result into exploitation pressure that has been there

40

Tanz. J. Sci. Vol. 44(3) 2018 over the years than the intensively protected community distribution and this agrees with Zaraninge Forest that had stable plant neutral diversity theory of species dispersion communities. Selective exploitation of (Hubbel 2001). Disturbance triggered the woody species creates large canopy gaps for recovery in favour of opportunistic plant light to reach the forest floor, a condition species in Kazimzumbwi and Pande forests that favored establishment of pioneer species that increased richness than low species (Ndangalasi 1997). Trema orientalis and richness in the pristine conditions of Panicum trichocladum are among species Zaraninge Forest. The undisturbed condition benefited from the gaps created through in Zaraninge forest made a few plant species disturbance within these forests (Mligo including Scorodophloeus fischeri, 2010). The aforementioned two species have Cynometra webberi, Cynometra invaded the exposed soil and can perform brachyrrachis, Brachystegia spiciformis and well in soils with poor fertility. Tessmannia burtii to forms a climax community (Mligo et al. 2009). The Variation in plant species richness among aforementioned species are better adapted selected forests under pristine conditions and have Regardless of being the least disturbed, outcompeted other species because of their Zaraninge Forest had low species richness ability to exploit large amount of than the disturbed Kazimzumbwi and Pande environmental resources and consequently Forests (Table 1). Disturbance in the later low in species richness as pointed out by forests were of different forms, while some Luoga (2000). parts within the forest have been selectively exploited of large canopy plant species, Variation in plant species evenness among some parts were heavily affected. The selected forests response of species to disturbance resulted The role of species evenness in biological into difference in richness among forests. If conservation is very important, although it is the frequency and magnitude of disturbance rarely examined (Mattingly et al. 2007). It increase beyond the irreparable level, only was recorded higher plant species evenness the colonizing species with high growth in Pande Forest than Zaraninge and rates and dispersal rates are able to exist in a Kazimzumbwi Forests with no significant plant community (Moshi 2000). However, difference among them. However, evenness the decrease in plant species richness may in the study forests was generally lower than constitute a temporary ecosystem change in savanna ecosystems (Mligo 2006). Low (Halpern and Spies 1995). That means plant evenness indices are indications of communities in the affected coastal forests inequitable distribution pattern among may recover through natural regeneration of individuals of species within plant indigenous plant species provided that there communities in the forests studied. Terazaki is an improved conservation management (1999) pointed out that plant community (Mligo et al. 2011). This is true in the with evenness indices less than 0.588 is disturbed Pande and Kazimzumbwi Forests unstable and communities with evenness that were dominated by colonizers following indices above 0.851 are indications of high selective removal of woody species that ecosystem stability. The evenness indices gave room to understorey and ground layer from this study implies that vegetation species to regenerate successfully that communities in the coastal forests are less contributed to high species richness. A few stable and can change rapidly through any small plant populations disperse seeds that form of disturbance. The lack of significant successfully germinates in a new discrepancy of the overall species evenness environment (disturbed area) predict plant among forests studied is because of being

41

Mligo - Application of regression model to identify a parameter that best defines species exposed to similar environmental conditions diversity curve displayed similar pattern influenced by the Indian Ocean. However, with that of species evenness, but deviated habitat fragmentation has contributed to substantially from that of richness because uneven species distribution and the two parameters have no significant consequently low evenness among relationship (Figure 3). Hill (1973) pointed individuals of plant species within these out that richness regulates variation in both forests. Some species have individuals evenness and diversity, however scaled widely distributed among habitats or down by Wisley and Stirling (2007) that, restricted to only one or two habitats and a species evenness predicts large proportions number of species in coastal forests fall of covariations among the three ecological under later category. For example parameters which is in agreement with the Scorodophloeus fischeri, Cynometra current observation. The weak positive webberi, Cynometra brachyrrachis and correlation between species richness and Tessmannia burttii are the dominant canopy evenness in each forest implies that the two tree species in Zaraninge forest (Mligo et al. parameters provide different interpretation 2009, 2010), whereas a pronounced on the coastal forests’ biodiversity. dominancy of Spirostachys africana exists The examined balance between the two in Pande Forest. Only a few pure stands of ecological parameters (richness and Manilkara sulcata were found in the evenness) using the linear model indicated southern and western parts in Kazimzumbwi that evenness parameter significantly forest with a well identifiable zonation of determines change in species diversity than plant communities in this forest such as the richness. The species evenness therefore hilltops (Manilkara sulcata), hill slopes provides adequate interpretation of change (Millettia puguensis and Scorodophloeus in plant biodiversity within coastal forests fischeri) and or valley bottoms (Antiaris with characteristic higher gradients than toxicaria and Milicia excelsa). Therefore, richness that had insignificant influence. low species evenness in this forest can be This observation can be strongly supported contributed by complex habitat zonation by Triin et al. (2009), that evenness accounts existing in the forest. for the major variation in the species diversity on composite vegetation data from The definition of plant species diversity different study areas whereas species based on linear model richness contributes less. It is therefore Data from coastal forests of Tanzania rejected the assumption that both evenness (Burgess and Clarke 2000, Howell et al. and richness have equal influence on species 2009, 2011), are usually presented in terms diversity and this gives justification that of number of species (richness) in a species evenness can be used as an adequate particular fragment excluding other forms of measure of plant biodiversity characteristics species assemblages. Stirling and Wilsey of the coastal forest ecosystem. (2001), Purvis and Hector (2000) pointed out that using species richness alone is However, the application of evenness insufficient parameter to describe plant concept to determine the plant biodiversity community characteristics without putting conservation value is likely to vary from one into consideration of species evenness. ecosystem to another. This is because the However the established definition of Eastern African coastal forest ecosystem is species diversity by Kent and Coker (1992) composed of large number of plant species and Magurran (2004) included both richness with habitat restrictions (endemism) that and evenness of species and this may have contributed to low evenness ecologically sounds. The established species indices. Low evenness indices influenced

42

Tanz. J. Sci. Vol. 44(3) 2018 the sensitivity of the regression model to provides a true picture of the ecological identify its commanding power on variation characteristics of the coastal forests in in species diversity within these forests. It Tanzania. The way individuals of species are was predicted that species habitat distributed within any particular plant fragmentation and local endemism are a community determines the means through function of evenness since many individuals which an ecosystem is defined. Therefore, of plant species restricted within one or a plant species evenness usually was an few plant communities within coastal adequate predictor of the coastal forests’ forests. It is therefore, the evenness conservation value because coastal forests component that commands the strongest have a large number of plant species with measure of diversity because of being limited distribution patterns. Designing and associated with the level of management and implementation of forest biodiversity habitat conditions than richness measures conservation strategies should be regional which is associated with natural regeneration specific and should consider fragments that potentials of the indigenous plant species. are in close proximity within an ecological Therefore the model is suitable to predict the landscape. This will guarantee conservation trend of change in plant species diversity of many species that are localized within indirectly when anthropogenic or other coastal forest fragments in Tanzania. forms of disturbance affect species evenness within a forest. This tells that plant species ACKNOWLEDGMENTS evenness is more sensitive and responds The author is very grateful of the support faster under different forms of disturbance from WWF through CEPF fund that than richness. The model usually is supported part of the fieldwork. Gratitude is consistent in vegetation data collected from extended to Mr. Haji Suleman a botanical the same phytogeographical region. Also the specialist in the identification of plants in the sampling technique should not be based on a field and specimen in the herbarium. The frequency measures because the widespread author is indebted to colleagues for species can be underestimated and rare comments before submission of the species can be overestimated which may not manuscript. reflect the importance of the evenness parameter as a measure of the conservation REFERENCES value of an ecosystem. Alatalo RV 1981“Problems in the measurements of evenness in ecology”. CONCLUSION Oikos 37: 199-204. It was found significant difference in plant Bock CE, Jones ZF and Bock JH 2007 species diversity among coastal forests Relationships between species richness, because of differences in conservation evenness, and abundance in a management regimes. The linear model southwestern savanna. Ecol. 88:1322– showed plant species diversity was 1327. influenced by richness and evenness at Burgess ND and Clarke GP 2000 Coastal different levels and this implies that the two Forest of Eastern Africa. IUCN-The parameters provide two different ecological World Conservation Union. Gland, interpretations of the conservation value of Switzerland and Cambridge, U.K. an ecosystem. The trend line for species Burgess ND, Clarke GP, Madgewick J, diversity curve follows a similar pattern as Robertson S A and Dickinson A 2000. the trend line of species evenness, means Distribution and Status. In. Burgess ND that, in any case species evenness is an and GP Clarke, (eds).The Coastal ecological parameter that adequately Forests of Eastern Africa, IUCN.

43

Mligo - Application of regression model to identify a parameter that best defines species

Clarke G P and Burgess ND 2000 Geology of tropical moist forest in Kibale Forest and geomorphology, In: Burgess ND and Reserve, Uganda. Unpublished PhD Clarke GP (eds). Coastal Forests of Thesis, Michigan State University, Eat Eastern Africa, pp.29-39. Norwich Ltd Lansing, USA. UK. Kent M and Coker P 1992 Vegetation Clarke G P and Dickinson A 1995 Status Description and Analysis. A practical report for 11 coastal forests in the coast Approach. John Willey and Sons, New region of Tanzania. Frontier-Tanzania York, pp. 319. Technical Report No.17. The Society for Koné B, Bognompke JP, Cherif M, Oikeh Environmental Exploration, UK/ S, Konan F and Traoré M J 2014 University of Dare es Salaam, Tanzania. Characterization of Soil Nutrients Frazier J G 1993 Dry coastal ecosystems in Leaching in Perched Groundwater as Kenya and Tanzania. In Van der Maarel, Occurring in Bush Fallow of Forest E (ed.) Ecosystems of the World 2B- Dry Zone. 1nt. J. Ecol. Env. Sci. 40 (2-3): Coastal Ecosystem: Africa, America, 165-174. Asia, and Oceania. Elsevier, Amsterdam. Luoga EJ 2000 The effects of Human Gaston K J and Spicer J I 1998 disturbance on diversity and dynamics of Biodiversity: An introduction. Blackwell the Eastern Tanzania miombo Science, Oxford, pp.420. arborescent species, PhD. Thesis. Graphpad Instat 2003. GraphpadInstat Witwatersrand University. Johannesburg, software Inc, 3.06. South Africa. Halpern CB and Spies TA 1995 Plant Ma M 2005 “Species richness vs evenness: species diversity in natural and managed independent relationship and different forests of the Pacific Northwest. Ecol. responses to edaphic factors”. Oikos 111: Appl. 5: 913-934. 192–198. Hamilton AJ 2005 Species diversity or Magurran A E 2004 Measuring biological biodiversity? J. Env. Manag. 75: 89-92. diversity. Blackwell Science, Oxford. Heyer WR Donnelly MA, McDiarmid RW, Mattingly WB, Hewlate R and Reynolds Hayer LC and Foster S eds. 1994 HL2007 “Species evenness and invasion Measuring and Monitoring Biological resistance of experimental grassland Diversity: Standard Methods for communities”. Oikos 116:1164–1170. Amphibians. Smithsonian Institute Press, Mligo C, Lyaruu HVM and Ndangalasi H J Washington D.C. 2011 The Effect of Anthropogenic Howell KM, Msuya CA, Mligo C, Werema disturbances on Population structure and C, Kihaule P, Honorati NK and regeneration potentials of Suleiman HO 2012 Biodiversity surveys Scorodophloeus fischeri and Manilkara of poorly Known Coastal Forests of sulcata in coastal forests of Tanzania. J. Southern, Eastern Tanzania and For. Sci. 73(1): 33–40. Zanzibar. Mligo C, Lyaruu HVM, Ndangalasi H J and Hill MO 1973 Diversity and Evenness: A Marchant R 2009 Vegetation community unifying notation and its consequences. structure, composition and distribution Ecol. 54: 427-432. pattern in the Zaraninge Forest, Hubbel SP 2001 The unified neutral theory Bagamoyo District, Tanzania. J. East of biodiversity and biogeography. Afr. Nat. Hist. 98 (2): 223–239. Princeton University Press. Mligo C 2006 Effects of Grazing Pressure Kasenene JM 1987 The influence of the on Plant Species composition and mechanized, selective logging, felling Diversity in the Semi-Arid Rangelands intensity and gap size on the regeneration

44

Tanz. J. Sci. Vol. 44(3) 2018

of Mbulu District, Tanzania. Agr. J. 1 Stohlgren TJ, Falkner MB and Schell LD (4): 277-283. 1995 “A Modified-Whittaker nested Mligo C 2010 Composition gradient in vegetation sampling method”. Vegetatio species distribution and genetic diversity 17: 113-121. of Scorophloeus fischeri Taub. In coastal Terazaki M 1999 “Mass Occurrence of forests, Tanzania. PhD thesis, University Bathypelagic Chaetognath Eukrohnia of Dar es Salaam. fowleri from the Arabian Sea and Bay of Mligo C 2015 Plant Species Composition Bengal”. Ind. J. Mar. Sci. 28: 163-168. and Distribution in Relation to Land Use Triin R, MartinT S, Lotten JJ, Mikael L, Patterns in Serengeti ecosystem Karin H, Marie V and Honor CP 2009 Tanzania. Op. J. For. 5: 607-620. “Small-scale plant species richness and Moshi GY 2000 The influence of evenness in semi-natural grasslands Disturbance on the status and respond differently to habitat Regeneration rate of selected Forests in fragmentation” Biol. Cons. 142: 899- the Udzungwa Mountains, Tanzania. 908. M.Sc. Thesis, University of Dar es Watanabe NM and Suzuki E 2008 Species Salaam. diversity, abundance, and vertical size Myers N, Mittermeier RA, Mittermeier CG structure of rattans in Borneo and Java. da Fonseca GAB and Kent J 2000 Biod. Cons. 17: 523–538. “Biodiversity hotspots for conservation White F 1983 The vegetation of Africa. priorities” Nat. 403: 853-858. Natural Resource Research, No. 20. Ndangalasi HJ 1997 Canopy gap Report to accompany characteristics and regenerating species UNESCO/AETFAT/UNSO Vegetation in Pugu Forests Reserve, Tanzania. Map of Africa, UNESCO, Paris. M.Sc. Thesis, University of Dar es Wilder C, Brooks T and Lens L 1998 Salaam. “Vegetation structure and composition of Purvis A and Hector A 2000 “Getting the the Tahita Hills Forests”. J. East Afr. measure of biodiversity”. Nat. 405: 212– Nat. Hist. 87: 181-187. 219. Wilsey B and Stirling G 2007 Species Shannon CF and Weiner W 1948 The richness and evenness respond in a Mathematical Theory of different manner to propagule density in Communication. University of Illinois developing prairie microcosm Press, Urbana. pp150. communities. Plant Ecol. 190:259–273. Stirling G and Wilsey B 2001 “Empirical WWF 2002 “Ecoregion Profile: Eastern Arc relationships between species richness, Forests. Tropical & Subtropical Moist evenness and proportional diversity”. Broadleaf Forests”. Cons. 5: 431-433. Am. Nat. 158:286–299.

45